See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/272150171

Digital watershed representation within the NewAge-JGrass system

Article · July 2014

CITATION READS

1 135

5 authors, including:

Giuseppe Formetta Andrea Antonello

Centre for Ecology & Hydrology Libera Università di Bozen-Bolzano
40 PUBLICATIONS   185 CITATIONS    20 PUBLICATIONS   106 CITATIONS   

SEE PROFILE SEE PROFILE

Silvia Franceschi Olaf David

Libera Università di Bozen-Bolzano Colorado State University
14 PUBLICATIONS   80 CITATIONS    79 PUBLICATIONS   587 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Spatial Measurement and Modeling of Agricultural Watersheds View project

http://abouthydrology.blogspot.com View project

All content following this page was uploaded by Riccardo Rigon on 17 September 2015.

The user has requested enhancement of the downloaded file.

Formetta, G., Antonello, A., Franceschi, S., David, O., Rigon, R. 2014. Digital watershed representation within the NewAge-JGrass system. Boletín Geológico
y Minero, 125 (3): 371-381
ISSN: 0366-0176

Digital watershed representation within

the NewAge-JGrass system
G. Formetta(1), A. Antonello(2), S. Franceschi(2), O. David(3), R. Rigon(4)
(1) Unical, University of Calabria, 87036 Arcavacata di Rende (CS)
(2) HydroloGIS Environmental Engineering, 39100 Bolzano - Italy
(3) Dept. of Civil and Environmental Engineering, Colorado State University, Fort Collins, CO, USA
(4) University of Trento, 77 Mesiano St., 38123 Trento, Italy

ABSTRACT

This paper describes the theoretical and technical architecture of a digital watershed model (DWM) that can
be used as a basis for storing data relative to a river basin and to run a spatially distributed hydrological mo-
del that is directly representable in a GIS (such as uDig). The DWM deployment is based on the integration of
geospatial modelling libraries (the Geotools) by means of which geographical information is encoded and be-
comes an integral part of the modelling process. Subsequently, the paper identifies a sequence of operations
that have to be performed in order to obtain the desired DWM, and briefly describes the necessary programs.
For the topological ordering of the network, a generalisation of the Pfafstetter ordering scheme is presented
as well as a partitioning scheme based on the distances to the divides (namely Hack’s Lengths).
These programs are stand-alone components which can be run in different ways by using: the uDig Spatial
Toolbox interface, the user friendly Object Modelling System (OMS) console or by command line. The OMS
console completes the work by allowing the connection of all parts to produce well-designed modelling solu-
tions in which the components are connected and executed together. The tools presented have been applied
to the Little Washita river basin (Oklahoma, US) and to the Piave river basin (Veneto, Italy).

Key words: digital watershed, geotools, Hack’s Lengths, Pfafstetter ordering.

Representación digital de cuencas hidrográficas dentro del sistema

NewAge-JGrass
RESUMEN

Este trabajo describe la arquitectura teórica y técnica de un modelo digital de cuenca (MDC) que puede uti-
lizarse para almacenar datos relativos a una cuenca fluvial y ejecutar modelos hidrológicos espacialmente
distribuidos que son directamente representables en un Sistema de Información Geográfica (como uDig). La
estructura del MDC está basada en la integración de librerías de modelado geoespacial (los Geotools) por
medio de las cuales la información geográfica es codificada para ser parte integral del proceso de modelado.
Subsecuentemente el trabajo identifica una secuencia de operaciones que tienen que realizarse para obtener
el MDC deseado, y describe brevemente los programas necesarios para obtenerlo. Para el ordenamiento
topológico de la red, se presenta una generalización del esquema de ordenación de Pfafstetter así como un
esquema de particionado basado en las distancia a las divisorias (las longitudes de Hack).
Estos programas son componentes independientes que se pueden ejecutar de varios modos: la interfaz
Spatial Toolbox de uDig, el modo amigable de consola Object Modelling System (OMS) o por comandos en
línea. La consola OMS completa el trabajo permitiendo la conexión de todas las partes para producir solu-
ciones de modelado bien diseñadas en las que los componentes están conectados y se ejecutan juntos. Las
herramientas presentadas se han aplicado a la cuenca fluvial Little Washita (Oklahoma, EEUU) y la cuenca del
río Piave (Veneto, Italia).

Palabras clave: cuenca digital, geotools, logitudes de Hack, ordenamiento de Pfafstetter.

371
Formetta, G., et al. 2014. Digital watershed representation within the NewAge-JGrass system. Boletín Geológico y Minero, 125 (3): 371-381

VERSIÓN ABREVIADA EN CASTELLANO

IIntroducción y metodología

La definición e implementación de modelos digitales de cuenca (MDC) está adquiriendo cada vez más impor-
tancia en hidrología especialmente con la introducción de las dimensiones espaciales en el modelado hidro-
lógico. Un MDC, como representación electrónica de las características hidrológicas espaciales y de series
hidrológicas temporales, tiene que operar con diferentes tipos de datos: elevación, características del agua,
uso del suelo, datos de observación puntual, y/o datos sobre una malla regular (imágenes de satélites, pro-
ductos climáticos, …). En este contexto los datos están relacionados y son utilizables para investigaciones en
un ambiente participatorio para promover una amplia colaboración entre muchos tipos de científicos e inge-
nieros (http://www.cuahsi.org).
Este trabajo presenta la arquitectura y la informática de un MDC capaz de almacenar datos relativos a una
cuenca fluvial y ejecutarse sobre un modelo hidrológico espacialmente distribuido. Todos los elementos se
pueden implementar y representar en un Sistema de Información Geográfica (GIS) de dominio público y có-
digo abierto, uDig que corre como cliente bajo diferentes plataformas (Windows, Mac OS/X, and Linux) y está
orientado como servicio web.
El nuevo MDC es una malla especial sobre la que el usuario puede ejecutar el modelo hidrológico semi-
distribuido NewAge-JGrass, un modelo continuo para predecir y modelar los recursos hídricos en general a
escala de cuenca. Se integraron diferentes componentes hidrológicos en el sistema tales como diferentes ti-
pos de algoritmos de interpolación meteorológica, balances energéticos de onda corta y onda larga, fusión de
la nieve y modelado de la escorrentía. Por estas razones el modelo hidrológico semi-distribuido NewAge re-
presenta un intento para representar y modelar todos los procesos hidrológicos. El diagrama de flujo de tra-
bajo comienza con la delineación de subcuencas utilizando las herramientas disponibles en el SIG uDig y los
JGrassTools. El sistema se basa en una partición geométrica del paisaje en base a laderas con enlaces. La
unidad básica para la evaluación del balance de agua es la ladera, las cuales se tratan como cajas negras. Sin
embargo, las laderas puede diseccionarse todavía más dependiendo de los procesos a analizar. Cada ladera
drena en un enlace simple asociado en lugar de en celdas o píxeles. Los canales se describen como elementos
vectoriales que están topológicamente interconectados en un grafo orientado simple. Además cada elemento
de la red del río puede incluir estructuras antrópicas que regulan los regímenes de flujo, con lo que se hace
posible simular entradas, gestión de presas, canales artificiales y extracciones de agua por ejemplo para rie-
go. Esta partición conceptual se desarrolló utilizando una informática con estructura vectorial para los canales
y estructura ráster para las laderas.
El artículo ilustra el procedimiento completo que el usuario ha de llevar a cabo para obtener su propio
MDC, esto es, la estructura ladera-enlace de la cuenca. Se ha esquematizado en la figura 1. Para el ordena-
miento topológico de la red, se presenta una generalización del esquema del ordenamiento de Pfafstetter así
como un esquema de particionado basado en las distancias a las divisorias (longitudes de Hack).
Las herramientas presentadas se han aplicado a dos casos de estudio, la cuenca del río Little Washita
(Oklahoma, EEUU) y la cuenca del río Piave (Veneto, Italia) que presentan diferencias en tamaño y en comple-
jidad topográfica.

Resultados y discusión

El trabajo discute el problema de la definición de un modelo digital de cuenca (MDC) en el contexto del siste-
ma hidrológico NewAge-JGrass. El procedimiento para obtener el MDC se resume en dos pasos i) análisis
geomorfológico de la cuenca con herramientas integradas en NewAge-JGrass y ii) ordenación y clasificación
de las subcuencas. El flujo de trabajo del primer paso se presenta en la figura 2 e incluye: i) relleno de todas
las posibles depresiones en algunos de los puntos del modelo digital de elevaciones; ii) cálculo de la dirección
de drenaje; iii) extraer la red de canales de drenaje; iv) etiquetado utilizando el mismo número para cada canal
y ladera que drena a dicho canal; v) cálculo del fichero shape que contiene información acerca de cada canal
y enlace tal como: área, perímetros, elevación media, etc.
La segunda parte está relacionada con el ordenamiendo de canales y subcuencas de la cuenca vertiente.
Una versión modificada del esquema de ordenamiento de Pfafstetter se implementó y chequeó en dos cuen-
cas de diferente tamaño. El procedimiento para calcular el ranking de la red de canales incluye: i) el cálculo de
la distancia de Hack (la distancia desde la cuenca a lo largo de la red procediendo aguas arriba a lo largo de la
línea de máxima pendiente); ii) cálculo de cauces Hack que numera una red de canales comenzando por el
canal principal y finalmente iii) aplicación de la metodología Pfafstetter. El algoritmo es capaz de gestionar y
etiquetar de un modo apropiado tanto lagos como infraestructuras antrópicas en la cuenca fluvial como pan-
tanos y derivaciones. Los resultados de cada uno de los programas para las cuencas fluviales de Little Washi-
ta y Piave se presentan en las figuras 4, 5 y 6. Los MDCs estructuradas en laderas y enlaces serán la base de
los componentes hidrológicos de NewAge-JGrass. Cada ladera y enlace se puede relacionar a ficheros CSV

372
 Formetta, G., et al. 2014. Digital watershed representation within the NewAge-JGrass system. Boletín Geológico y Minero, 125 (3): 371-381

conteniendo datos hidro-meteorológicos tales como lluvia media y temperatura media del aire para laderas,
así como descarga media y velocidad media para canales. Sin embargo, está claro que el sistema se puede
expandir para comunicarse fácilmente con la base de datos y servidores, de tal modo que se podrían escalar
a situaciones operacionales reales donde tales sistemas son necesarios.
Las salidas de las herramientas es un modelo de datos implementado utilizando características simples
OGC, esto es, ficheros shape e información conectada a ellos. Cada herramienta es un componente, de acuer-
do a las definiciones en Sistema de Modelado de Objetos (SMO), esto es, un entorno de modelado basado en
Java desarrollado para modelado por componentes. Se basan en Geotools, librerías geoespaciales por medio
de las cuales se codifica información geográfica. Las herramientas mismas se pueden ejecutar por ellas mis-
mas o se pueden conectar fácilmente con otros componentes OMS que construyen una cadena de modelos.
Todos los programas presentados, así como los componentes NewAge, se pueden ejecutar de tres modos
diferentes: la interfaz Spatial Toolbox uDig, presentada en la figura 3, la consola amigable SMO o la tradicional
línea de comandos. En los primeros dos casos, una interfaz gráfica de usuario facilita el procedimiento de
entrada-salida para las aplicaciones de las herramientas y los resultados se pueden visualizar fácilmente en el
SIG uDig mediente la selección y arrastre de mapas en el panel de vista. Es más, uDig es capaz de tomar geo-
datos servidos a través de estándares como el Web Feature Services (WFS), Web Map Services (WMS), y el
Web Coverage Services (WCS). Esto posibilita el hecho de que la información topológica y geométrica relacio-
nada con la cuenca fluvial se puede traducir muy fácilmente en tablas relacionadas de un geo-database SQL.
Esta comunicación se realizó en una implementación piloto para la Autoridad de la Cuenca del Río Adige,
donde se modelo la cuenca fluvial del río Adige.
Para concluir, el trabajo presenta las herramientas y el procedimiento para obtener un MDC que puede ser
usado como dominio espacial para la ejecución de modelos hidrológicos. En particular la ordenación Pfafstet-
ter de la red se modifica y se aplica en dos casos de estudio. Permite una fácil navegación a través de la red
fluvial, ayuda al usuario a determinar si dos cauces de río están conectados y cómo están conectados y final-
mente permite analizar la cuenca a diferentes escalas de resolución espacial de acuerdo al objetivo del análi-
sis desarrollado.

Introduction the Consortium of Universities for the advancement

Even though highly remarkable results have been
obtained in hydrological science by using just point
models (e.g Rodriguez-Iturbe and Porporato (2005),
Eagleson (1970)), introducing spatial dimension in hy-
drological modelling is the source of more physical
realism and of new dynamical features. One of the
goals of NewAge-JGrass (Formetta et al., (2011)) is to
advance this philosophy and to allow the construction
of minimalist model solutions (MS) that are aware of
the spatial information and its heterogeneity when
they are used for predictions.
Using this modelling strategy however, it is ne-
cessary to conceptualise basin description in formal
terms, usually called the digital watershed model
(DWM). A DWM is an electronic representation of a
watershed’s spatial characteristics and time-series
hydrologic information. “It can include elevation,
water features, land use, point observation data,
and/or gridded data, e.g. remote sensing, climate
products, where the data are related and usable for
investigations in a participatory environment to pro-
mote a broad collaboration among many types of
scientists and engineers” (from: http://www.cuahsi.
org). One of the most mature digital watershed de-
signs of schematisation is the one initially encoded
by Maidment (2002), and subsequently endorsed by
of Hydrological Sciences (CUASHI), and described
for instance by Goodall and Maidment (2009). It dis-
tinguishes the basic units that make up a watershed
and identifies an appropriate data model and storage
formats for them. For instance, Arc Hydro, Maidment
(2002), distinguishes the following as part of the ba-
sin:
t

UIF
TUSFBN
OFUXPSL
TVCEJWJEFE
JO
MJOLT
t

UIF
CBTJOT
TVCEJWJEFE
JO
TVC
CBTJOT
t

UIF
MBLFT
BOE
TVSGBDF
XBUFS
CPEJFT
t

UIF
NPOJUPSJOH
QPJOUT
The Arc Hydro data model also includes the ground-
water features and dynamics (Strassberg et al., (2011))
that are not covered here. DWMs are usually stored
as simple features according to the Open GIS Con-
sortium (OGC) standard that can be processed by all
major GISs and have a corresponding storage format
within all major database systems that manage geo-
graphic features.
The effective use of spatial information in models
usually requires a further refinement, which is very
much modelling dependent, and is usually based on
the definition of the hydrologic response units (HRU)
Ross et al., (1979), Flugel (1995), Krause (2002) and
Viviroli et al., (2009). These are the elementary parts of
the basins that are treated as black box units. Moussa
et al., (2007) represents an example of a detailed par-

373
Formetta, G., et al. 2014. Digital watershed representation within the NewAge-JGrass system. Boletín Geológico y Minero, 125 (3): 371-381
tition of a basin for agricultural use. Within the AGE
model (Ascough II et al., (1999)) sub-catchments are
subdivided into many detailed functional parts which
refer to the different treatments in soil use and/or
land-cover which are thought to influence hydrologi-
cal fluxes.
In this paper we describe the digital watershed
model, i.e. the modern data model, used in the
NewAge JGrass modelling system. The DWM is ai-
med to be a flexible and efficient description of the
spatial features of a basin. It is needed to drive spa-
tially explicit hydrological models (or model com-
ponents) and is able to include human infrastructu-
res inside the river basin description, in order to
account properly for anthropic alteration of the hy-
drologic cycle. Section 2 introduces the hydrologi-
cal components of the NewAge-JGrass model and
explains how they fit into the digital watershed mo-
del implemented into the system. Sections 3 and 4
introduce the method and the tools available in the
system for performing the basin geomorphologic
analysis and the consequent watershed delineation
according a suitable modification of the Pfafstetter
ordering scheme. Finally, two applications of the
method are presented
Concepts
In NewAge-JGrass (Formetta et al., (2011), Formetta et
al., (2013), Formetta et al., (2014a) and Formetta et al.,
(2014b)), the basin is partitioned into a hillslope-link
(HL) structure, where the hillslopes are the basic hy-
drologic response units (HRU) for all that concerns
rainfall-runoff. Channels are described as vector fea-
tures that are topologically interconnected in a sim-
ple, directed graph. For computational reasons, parti-
tioning of the area is not usually designed to identify
all the physical hillslopes present in the system, but to
define the dimensions of small watersheds. In the
applications shown in this paper, these are of 2-10
Km2 on average. HRUs can either be represented as
vector features or rasters.
Within a model, any element of the river network
can include anthropogenic structures that regulate
the flow regimes, thus making it possible to simulate
intakes, management of dams, artificial channels, and
water abstractions, for example, for irrigation.
Hillslopes can be further dissected depending on
the processes to be analysed. For instance, when
temperature is the concern, each hillslope can be fur-
ther subdivided into altimetric bands, each one with
its own temperature, that could eventually be mani-
pulated to obtain a single value for the whole hillslope,
374
depending on the requirements of the case in study.
Regarding the estimation of radiation, or snow, the
specific model components of NewAge-JGrass can
either use information at pixel scale, which is sub-
sequently averaged, or information from selected
representative points within the hillslope. Therefo-
re, besides, a generic delineation of the basin, each
module component relative to a process can use
the data and the geometries it requires, and in ge-
neral the HL partition is a key for interrogating and
treating other informative layers. One critical issue
is how single units exchange the main hydrological
fluxes. MYDHAS [Moussa et al., (2010)] and AGEs
(Ascough II et al., (1999). HRUs, for instance, ex-
change runoff and subsurface fluxes in multiple di-
rections and therefore have procedures to manage
this complexity. NewAge-JGrass, at the present sta-
ge, just allows one hillslope to discharge into its
channel link. In any case, the river network constitu-
tes a hierarchy in which sources flow into the inter-
nal links and these, in turn, into larger streams. To
account for this hierarchical simulation various
strategies can be used. The most advanced scheme
is probably Liu et al., (2013), which builds on the
knowledge obtained by analyzing the Strassberg et
al., (2011) generalised Pfafstetter’s scheme that is
used here, and is described below.
Whatever the conceptualisation, the challenge is
to deploy the ideas in a robust, correct code. This is
accomplished in NewAge-JGrass by using the
Geotools libraries and their implementation of the
geographic features which seamlessly integrate with
the Object Modelling System version 3 (OMS3) (Da-
vid et al., (2013) and uDig (http://udig.refractions.
net/).
To obtain this hierarchical structure it is first ne-
cessary to process the raster data from a digital ele-
vation model. For the scope of building the DWM,
this paper also covers the description of the main
tools for basins and channels (river networks),
which are included in the Horton Machine Rigon et
al., (2006) and Abera et al., (2014), and serve as
foundations. These have been evolved during the
last twenty years in parallel to other suites of pro-
grams such as Taudem (Tarboton (2005)) and River-
Tools (Peckham (2003)). In contrast, the Horton Ma-
chine (HM) is consistently integrated in a GIS system
through the uDig Spatial Toolbox. Formetta et al.,
(2014a) describes in detail how each command can
either work separately (from the command line or
the OMS3 console) or by being connected to other
commands through the functionalities offered by
the OMS3 system, thus offering different possible
operational modes.
 Formetta, G., et al. 2014. Digital watershed representation within the NewAge-JGrass system. Boletín Geológico y Minero, 125 (3): 371-381

Catchment analysis a point lower, and so it allows the identification

The analysis of the catchment starts with the acquisi-
tion of a Digital Elevation Model (DEM) of the catch-
ment, e.g. Wilson and Gallant (2000) and Peckham
(2003). It is performed as illustrated in Figure 1 and

BOE
.BSLT

summarized for the reader below.
The tools for basin characterization that the HM
(Abera et al., (2014)) offers are, amongst others:
t

3BTUFS3FBEFS
XIJDI
JNQPSUT
SBTUFS
NBQT
BTD

(D8); the input for the command is the output
.grass, .geotifi) and converts them into GridCove-
rage, the object used in Geotools to store raster
data.
t

7FDUPS3FBEFS
XIJDI
JNQPSUT
WFDUPS
åMFT
TIQ

flow. The deviation, calculated using a triangu-
and converts them into the Features Collection,
the object used in Geotools to store vector data.
t

1JUåMMFS
5BSCPUPO

of the drainage directions at each point; the in-
put for Pitfller is the DTM previously imported in
the GIS. An example of the output is shown in
Figure 2,a.
t

'MPX%JS
0$BMMBHIBO
BOE
.BSL

et al., (1984)): it calculates the drainage directions
with the method of the maximal steepest descent
slope, selecting one out of 8 possible directions
map of the Pitfiller component.
t

%SBJO%JS
UIJT
UPPM
QSPWJEFT
UIF
ESBJOBHF
EJSFD-
tions, minimising the deviation from the real
lar construction, could be given in degrees (D8
LAD method) or as transversal distance (D8 LTD


UIJT
UPPM
åMMT
UIF
EF- method), Orlandini et al., (2003) and Orlandini et
pression points in the DTM and assures that, for al., (2012). The input raster maps are: the map in
any point in the basin, except the outlet, there is output from Pitfiller and Flowdir. The outputs are

Figure 1. The workflow for basin characterization in NewAge-JGrass.

Figura 1. El flujo de trabajo para la caracterización de cuencas en NewAge-JGrass.

375
Formetta, G., et al. 2014. Digital watershed representation within the NewAge-JGrass system. Boletín Geológico y Minero, 125 (3): 371-381

the raster maps of the drainage directions, (Fig.
2,b), and of the total contributing area, (Fig. 2,c).
t

&YUSBDU/FUXPSL
JU
FYUSBDUT
UIF
DIBOOFM
OFUXPSL

from the drainage directions. Three operational
modes are implemented. They differ in the way
in which the start of the channel is modelled:
1. mode 0: by using a threshold value of the con-
tributing areas (then only the pixels with con-
tributing area greater than the threshold are
the channel heads);
2. mode 1: by using a threshold value of the pa-
rameter: equivalent to a threshold value of the
stress tangential to the bottom;
3. mode 2: by using a threshold value on the
stress tangential to the bottom;
After identifying the beginning of a channel, the
points downstream of it are considered as channel.
If “mode 0” is used, the inputs of ExtractNetwork
are the map outputs of Pitfiller and DrainDir
(both drainage directions and TCA). The output
will be the raster map and, if the user needs it,
also a vector file an output of the river network,
(Fig. 2,d ).
t

/FU/VNCFSJOH
JU
BTTJHOT
EJGGFSFOU
OVNCFST
UP

each network channel and labels it with the co-
rresponding hillslope number, which is connec-
ted to the link. The input maps are: the file contai-
ning the flow directions (generated by DrainDir)
and the map containing the channel network (ge-
nerated by extractnetwork). There are two output
raster maps: the network map with the numbe-
red streams and the map containing the labelled
sub-basins.
t

#BTJO4IBQF
JU
DPNQVUFT
UIF
TIBQF
åMF
PG
UIF
CB-
sin split into hillslopes. The BasinShape inputs are
the map output of Pitfiller and the map containing

Figure 2. The Little Washita basin: output of the Pitfiller Horton Machine (a), DrainDir, (b) TCA (c) and Extract Network (d).
Figura 2. La cuenca Little Washita: (a) salida de la Máquina Pitfiller Horton; (b) DrainDir; (c) TCA y (d) Extract Network.

376
 Formetta, G., et al. 2014. Digital watershed representation within the NewAge-JGrass system. Boletín Geológico y Minero, 125 (3): 371-381

the labelled sub-basins (output of NetNumbe-
ring). The resulting shape file contains the basin
split for each hillslope with some features such as:
– Area: the hillslope area [m2];
– Perimeter: the hillslope perimeter [m];
– Netnum: the hillslope ID;
– MaxZ: the hillslope maximum elevation [m];
– MinZ: the hillslope minimum elevation [m];
– AvgZ: the hillslope average elevation [m];
– Height: the hillslope centroid elevation [m];
Figure 3 presents a screenshot of the SpatialTool-
box-uDig GIS interface. An example of basin partition
in hillslopes is presented; for each hillslope the user is
able to visualize the geomorphological features such
as area, perimeter, mean elevation, etc.
Basin ordering and classification
In order to build the DWM it is necessary to assign
a topology to the river network. In the past, a few
methods for the topological ordering of the network
were presented and the original historically was the
Horton-Strahler [Horton (1945)]. This is the ordering
method implemented in the HM.
Strahler ordering was used in the original paper
about the geomorphologic unit hydrograph [Rodri-
guez-Iturbe and Valdes (1979)] but cannot be conside-
red still functional to modern semi-distributed mo-
dels, where the topography is assumed (and is) known
in great detail. Another topological partition of the
basin was based on magnitude ([Rodriguez-Iturbe
and Rinaldo (2001)] and this is also implemented in
the HM). However, magnitude was just a surrogate for
the total contributing area, when such a quantity was
difficult to estimate (on maps). More recent classifica-
tions of streams rely on other quantities, such as the
length from any point to the divides (e.g. Rigon et al.,
(1996) and Gangodagamage et al., (2011)) and on the
Pfafstetter numbering schemes [Verdin and Verdin
(1999)].
The Pfafstetter numbering scheme (PNS) algorithm
is defined as follows: starting from the outlet of the
watershed, the main stream is delineated first. It uses
the river streams extracted from the tree-like network
of the drainage directions according to the algorithm
presented in the previous section. This river network
presents links (channel segments) separated by
junctions, where tributaries meet. Each stream is cha-
racterized by a total upslope area, which is the total

Figure 3.The Little Washita basin split into hillslopes by using the SpatialToolbox and uDig.
Figura 3. La cuenca Little Washita dividida en laderas utilizando el SpatialToolbox y uDig.

377
Formetta, G., et al. 2014. Digital watershed representation within the NewAge-JGrass system. Boletín Geológico y Minero, 125 (3): 371-381

area of the basin flowing into that stream. The main
stream is obtained by following the river network,
starting from the outlet going upstream. When co-
ming to a junction, the direction of the stream follows
the channel link with the largest upslope area. In the
case of equal areas a random direction is chosen.
Each junction separates the main stream into links,
which are numbered with a series of odd numbers

JU
DBMDV-
starting with 1 at the outlet, (see Fig.5, below). Tributa-
ries of the main stream are numbered with increasing,
even, numbers while going upstream (assuming that
two tributaries do not flow into the main stream
at the same point), Figure 5. This generalises the
original Pfafstetter numbering where just 5 links and 4
tributaries were allowed to constitute the main river-
network tree in order to limit the use to 9 digits to
identify any link. Tributaries can have sub- tributaries.
As shown in Figure 5, one tributary has sub-tributa-
ries, and therefore a second order numbering is used,
represented by two digits separated by a point. The
main stream of 8 is split into links 8.1 and 8.3 and two
order 2 headwater basins are delineated: 8.2 and 8.3,
where 8.3 is thought to continue the main stream of
the tributary.
The commands to obtain the partition of the catch-
ment are therefore:
t

)BDL
-FOHUI
BU
B
HJWFO
QPJOU
JO
B
CBTJO
lates the distance from the watershed along the
network proceeding upstream along the maxi-
mal slope length. The input raster maps are: the
drainage directions map (obtained with DrainDir)
and the contributing areas map. The output is the
raster map of the Hack distances.
t

)BDL
4USFBN
JU
BSSBOHFT
B
DIBOOFM
OFUXPSL
TUBS-
ting from the main stream. The main stream is of
order 1 and its tributaries of order 2, the sub-tri-
butaries are of order 3, and so on. The input ras-
ter maps are: the drainage directions map (obtai-

Figure 4. The Little Washita basin: output of HackLength, (a), and Hack Stream, (b).
Figura 4. La cuenca Little Washita: (a) salida HackLength y (b) salida de Hack Stream.

378
 Formetta, G., et al. 2014. Digital watershed representation within the NewAge-JGrass system. Boletín Geológico y Minero, 125 (3): 371-381
ned with DrainDir), the total contributing areas,
the Hack length map (obtained with hack length),
and the channel network (obtained with extract
network).
t
5IF
1GBGTUFUUFS
BMHPSJUIN
CVJMET
UIF
UPQPMPHZ
PG

the network by numbering the river network
structure according to a generalisation of the Pfa-
fstetters numbering scheme (PNS) (e.g. Verdin
and Verdin (1999), Furst and Horhan (2009)).
Once the river network has been numbered accor-
ding to Pfastetter, it is naturally subdivided into parts,
if each channel is associated with its hillslope, so that
any part of the basin is uniquely identified. Besides,
PNS uniquely identifies the watershed channels
downstream of a point of interest (i.e. a droplet falling
into the 8.3 sub-catchment is guaranteed to flow into
links 8.1, 7, 5, 3, 1).
At the same time, it is known that, for instance, sub-
catchment 4.3 is neither upstream nor downstream of
8, but both merge into stream 3. A dense network or-
dering is presented in Figure 6, where the tools are
applied for the Piave river basin (Italy).
Figure 5. The Pfafstetter numbering scheme for the Little Washita
watershed, Oklahoma (U.S.). The figure shows an example of the-
Pfafstetter river-network numbering.
Figura 5. El esquema de numeración de Pfafstetter para la cuenca
vertiente de Little Washita, Oklahoma (EEUU). La figura muestra un
ejemplo de la numeración de Pfafstetter para la red de drenaje.
379
Conclusions
The paper introduces a set of tools for analysis and
ordering of channel networks from which a digital wa-
tershed model suitable for being used in spatially ex-
plicit hydrological modelling is obtained. The tool
outcome is a data model implemented by using OGC
simple features, i. e. shape files and information con-
nected to them. The tools themselves are implemen-
ted as OMS version 3 components, to be used within
the Spatial Toolbox of uDig. The conceptual core
around which these tools were built is the use of a
modified Pfafstetter numeration of the networks to ac-
count for the topology of the network and the basin.
This is particularly useful because it allows: (i) an easy
navigation through the river basin; (ii) to determine if
two links are connected (for example a link 8.3 is the
third link of the main stream of tributary 8); iii) to pru-
ne out the smaller channels to analyse the basin at
different scales of resolution, therefore being able to
deal with basins at multiple scales, according the sco-
pe of the analysis performed.
Figure 6. Pfafstetter numbering scheme for the Piave river basin
(Italy). The figure shows an example of the Pfafstetter river-network
numbering.
Figura 6. El esquema de numeración de Pfafstetter para la cuenca
del río Piave (Italia). La figura muestra un ejemplo de la numeración
de la red de drenaje de Pfafstetter.
Formetta, G., et al. 2014. Digital watershed representation within the NewAge-JGrass system. Boletín Geológico y Minero, 125 (3): 371-381
The more important tributary numbering is presen-
ted. PNS, therefore, supplies the topological informa-
tion that is otherwise missing in shape files represen-
ting the network.
Clearly, the above topological and geometrical in-
formation can easily be translated into interconnected
tables of an SQL database, as shown in a prototype
implementation for the River Adige Basin Authority. In
fact, the DWM described has been successfully used
in the JGrass-NewAGE model, of which a detailed re-
ference can be found in Formetta et al., (2011), For-
metta et al., (2014a), Formetta (2013) and Formetta et
al., (2013).
References
Abera, W., Antonello, A., Franceschi, S., Formetta, G., and
Rigon, R., 2014. The
uDig Spatial Toolbox for hydro-geomorphic analysis, Clarke,
L.E and Nield,
J.M. (Eds.) Geomorphological Techniques (Online Edition).
British Society for Geomorphology; London, UK. ISSN:
2047-0371.
Ascough II, J. C., Hoag, D. L., Frasier, W. M., and McMaster,
G. S., 1999. Computer use in agriculture: an analysis of
Great Plains producers, Computers and electronics in
agriculture, 23, 189-204.
David, O., Ascough II, J., Lloyd, W., Green, T., Rojas, K., Lea-
vesley, G., and Ahuja, L., 2013. A software engineering
perspective on environmental modeling framework de-
sign: The Object Modeling System, Environmental Mode-
lling & Software, 39, 201-213.
Eagleson, S., 1970. Dynamic hydrology, McGraw-hill Book
Company, 1970.
Flugel, W., 1995. Delineating hydrological response units by
geographical information system analyses for regional
hydrological modelling using PRMS/MMS in the draina-
ge basin of the River Brol, Germany, Hydrological Proces-
ses, 9, 423-436.
Formetta, G., 2013. Hydrological modelling with compo-
nents: the OMS3 NewAge-JGrass system, Ph.D.Thesis.
Formetta, G., Mantilla, R., Franceschi, S., Antonello, A., and
Rigon, R., 2011. The
JGrass-NewAge system for forecasting and managing the
hydrological budgets at the basin scale: models of flow
generation and propagation/routing, Geoscientific Model
Development, 4, 943-955, doi:10.5194/gmd-4-943-2011,
URL http://www.geosci-model-dev.net/4/943/2011/.
Formetta, G., Rigon, R., Chavez, J., and David, O., 2013. Mo-
deling shortwave solar radiation using the JGrass-NewA-
ge system, Geoscientific Model Development, 6, 915-928.
Formetta, G., Antonello, A., Franceschi, S., David, O., and
Rigon, R., 2014a. Hydrological modelling with compo-
nents: A GIS-based open-source framework, Environ-
mental Modelling & Software, 55, 190-200., 2014a.
Formetta, G., Kampf, S. K., David, O., and Rigon, R., 2014b.
Snow water equivalent modeling components in NewA-
380
ge-JGrass, Geoscientific Model Development, 7, 725{736,
doi:10.5194/gmd-7-725-2014, URL http://www. geosci-mo-
del-dev.net/7/725/2014/.
Furst, J. and Horhan, T., 2009. Coding of watershed and river
hierarchy to support GIS-based hydrological analyses at
different scales, Computers & Geosciences, 35, 688-696.
Gangodagamage, C., Belmont, P., and Foufoula-Georgiou,
E., 2011. Revisiting scaling laws in river basins: New con-
siderations across hillslope and fluvial regimes, Water
Resources Research, 47.
Goodall, J. L. and Maidment, D. R., 2009. A spatiotemporal
data model for river basin-scale hydrologic systems, In-
ternational Journal of Geographical Information Science,
23, 233-247.
Horton, R. E., 1945. Erosional development of streams and
their drainage basins; hydrophysical approach to quanti-
tative morphology, Geological Society of America Bulle-
tin, 56, 275-370.
Krause, P., 2002. Quantifying the impact of land use changes
on the water balance of large catchments using the J2000
model. Physics and Chemistry of the Earth, Parts A/B/C,
27, 663-673.
Liu, J., Zhou, Z., Jia, Y., Wang, H., and Chen, X., 2013. A stem-
branch-topological codification for watershed subdivi-
sion and identification to support distributed hydrologi-
cal modeling at large river basins, Hydrological Processes.
Maidment, D., 2002. Arc Hydro: GIS for water resources, vol.
1, ESRI press.
Marks, D., Dozier, J., and Frew, J., 1984. Automated basin
delineation from digital elevation data, Geo-processing,
2, 299-311.
Moussa, R., Chahinian, N., and Bocquillon, C., 2007. Distri-
buted hydrological modeling of a Mediterranean moun-
tainous catchment-Model construction and multi-site va-
lidation, Journal of Hydrology, 337, 35-51.
Moussa, R., Colin, F., Dages, C., Fabre, J., Lagacherie, P.,
Louchart, X., Rabotin, M., Raclot, D., and Voltz, M., 2010.
Distributed hydrological modelling of farmed catchments
(MHYDAS): assessing the impact of man-made structu-
res on hydrological processes.
O’Callaghan, J. F. and Mark, D. M., 1984. The extraction of
drainage networks from digital elevation data, Computer
vision, graphics, and image processing, 28, 323-344.
Orlandini, S., Moretti, G., Franchini, M., Aldighieri, B., and
Testa, B., 2003. Pathbased methods for the determination
of nondispersive drainage directions in grid-based digital
elevation models, Water resources research, 39, 1144.
Orlandini, S., Moretti, G., Corticelli, M. A., Santangelo, P. E.,
Capra, A., Rivola, R., and Albertson, J. D., 2012. Evalua-
tion of flow direction methods against field observations
of overland flow dispersion, Water Resources Research,
48.
Peckham, S., 2003. RiverTools Users Guide, Boulder, CO: Ri-
vix, LLC.
Rigon, R., Rodriguez-Iturbe, I., Maritan, A., Giacometti, A.,
Tarboton, D. G., and Rinaldo, A., 1996. On Hack’s law, Wa-
ter Resources Research, 32, 3367-3374.
Rigon, R., Ghesla, E., Tiso, C., and A, C., 2006. The Horton
machine: a system for DEM analysis : the reference ma-
nual.
 Formetta, G., et al. 2014. Digital watershed representation within the NewAge-JGrass system. Boletín Geológico y Minero, 125 (3): 371-381
Rodriguez-Iturbe, I. and Porporato, A., 2005. Ecohydrology
of water-controlled ecosystems: soil moisture and plant
dynamics, Cambridge University Press.
Rodriguez-Iturbe, I. and Rinaldo, A., 2001. Fractal river ba-
sins: chance and selforganization, Cambridge University
Press.
Rodriguez-Iturbe, I. and Valdes, J. B., 1979. The geomorpho-
logic structure of hydrologic response, Water Resources
Research, 15, 1409-1420.
Ross, B., Contractor, D., and Shanholtz, V., 1979. A finite-ele-
ment model of overland and channel flow for assessing
the hydrologic impact of land-use change, Journal of Hy-
drology, 41, 11-30.
Strassberg, G., Jones, N. L., and Maidment, D. R., 2011. Arc
Hydro Groundwater: GIS for Hydrogeology, Esri Press.
381
Tarboton, D. G., 1997. A new method for the determina-
tion of flow directions and upslope areas in grid digital
elevation models, Water resources research, 33, 309-
319.
Tarboton, D. G., 2005. Terrain analysis using digital elevation
models (TauDEM), Utah Water.
Verdin, K. L. and Verdin, J. P., 1999. A topological system for
delineation and codification of the Earths river basins,
Journal of Hydrology, 218, 1-12.
Viviroli, D., Zappa, M., Gurtz, J., and Weingartner, R., 2009.
An introduction to the hydrological modelling system
PREVAH and its pre-and post-processingtools,
Environmental Modelling & Software, 24, 1209-1222.
Wilson, J. P. and Gallant, J. C., 2000. Terrain analysis: princi-
ples and applications, Wiley.
Recibido:
Revisado:
Aceptado:
Publicado:
View publication stats